Koala species distribution and fertility-reduction model
Combining a point-process model, an ensemble species distribution model,
and a demographic model to project koala populations in the
Mount Lofty
Ranges over the next 25 years to assess the efficiency and
cost-effectiveness of fertility-control interventions.
Accompanies paper:
Saltré,
F, KJ
Peters, DJ
Rogers,
J
Chadœuf,
V
Weisbecker,
CJA
Bradshaw. Balancing overpopulation and conservation targets to
optimize koala management strategies. In review
Scripts & Prerequisites
Ensure the following R libraries are installed: geoR,
sf, spatstat, MASS,
proj4, sp, terra,
dplyr,
plotly,deSolve,biomod2,
terra, ggplot2, rgdal,
sf, raster, maptools
MLR Koala Population Projection Model
Overview
This script models the koala population in the Mount Lofty Ranges
using a matrix population model. It incorporates stochastic factors,
sterilization scenarios, and catastrophic events to project population
dynamics over the years 2016–2040. The model also assesses the
efficiency and cost-effectiveness of sterilization interventions.
Key Features
- Population Projection: Simulates future koala
population trends under unmanaged and sterilization scenarios.
- Stochastic Dynamics: Introduces randomness in
fertility and survival rates to account for real-world variability.
- Catastrophe Scenarios: Models catastrophic events
affecting koala survival.
- Sterilization Analysis: Evaluates the impact of
sterilization rates (0-90%) on population size and associated
costs.
- Cost Modelling: Estimates costs for sterilization
programs, considering population density and search efforts.
Outputs
- Population Projections: Median population size,
confidence intervals, and extinction probabilities for each
scenario.
- Sterilization Costs: Total and annual costs of
fertility-control programs.
- Population Metrics: CSV files summarizing
population size, sterilization counts, and intervention costs.
Species Distribution Model (SDM)
Overview
The Species Distribution Model (SDM) integrates
environmental and species presence/absence data to estimate habitat
suitability for koalas across the Mount Lofty Ranges.
Key Features
- Data Preprocessing:
- Cleans and filters species presence data (e.g., removes records from
Kangaroo Island).
- Projects species presence data to a consistent geographic coordinate
system.
- Environmental Data Integration:
- Loads and processes environmental predictors (e.g., distance to
roads, rainfall, vegetation cover).
- Combines all predictors into a raster stack for analysis.
- Checks for multicollinearity using the Variance Inflation Factor
(VIF).
- Modeling:
- Utilizes the
BIOMOD2 package to create individual
species distribution models.
- Implements cross-validation and evaluates variable importance.
- Ensemble Modeling:
- Combines individual models into an ensemble for robust
predictions.
- Produces evaluation metrics and variable importance plots for
ensemble outputs.
- Population Density Estimation:
- Converts habitat suitability scores into population density
estimates using regional calibration factors.
Outputs
- Habitat suitability raster maps.
- Evaluation metrics, including True Skill Statistic (TSS) and ROC
curves.
- Ranked variable importance plots.
- Projected population density maps, including confidence
intervals.
Script Details
- Input Data:
- Species data:
KoalaData(GKC1&2).csv
- Environmental layers: e.g.,
Dist2sealRoads.asc,
MinTemp.asc.
Point Process Model for Koala Spatial Analysis
Overview
This repository contains the implementation of a Point Process Model
designed for analyzing the spatial distribution of koalas in the Mount
Lofty Ranges. It integrates multiple spatial datasets and advanced
statistical methods to identify factors affecting koala intensity and
distribution. The model explores the influence of proximity to parks,
roads, and other geographic features. This Point Process Model: - Uses
spatial point pattern analysis to assess koala distributions. - Employs
kernel density estimation (KDE) to identify koala hotspots. - Fits
Poisson models to evaluate the effects of distance from hotspots, parks,
and roads. - Generates intensity maps for visualizing koala density and
habitat preferences.
Key Features
- Kernel Density Estimation (KDE): Highlights regions
of high koala intensity.
- Distance-Based Analysis: Explores the impact of
proximity to hotspots, roads, and parks.
- Poisson Models:
- Model 1: Effect of distance to hotspots.
- Model 2: Adds the influence of park boundaries.
- Model 3: Integrates distance to roads.
- Visualization:
- Intensity maps with and without park protection.
- Graphical representation of the influence of road proximity.
Outputs
- Koala Intensity Maps:
- Maps showing koala density across the study area.
- Separate visualizations for the impact of park protection and road
proximity.
- Statistical Summaries:
- Estimated koala density per square kilometer.
- Comparative intensity within and outside parks.
- Distance Effects:
- Plots illustrating how distance from hotspots and roads influences
koala intensity.